Export & import bundles

care export writes saved chains — and, optionally, the AgentSkills they depend on — into a single portable tarball you can copy to another machine. care import is its inverse: it reads chain JSON back and (with --apply) saves it into the local Memory. Both default to the latest channel. This pair is the terminal twin of the Library’s export bundle action.

care export <output> <entity_id>...

Bundle one or more chains into a tarball (e.g. bundle.tar.gz).

care export bundle.tar.gz weather-report risk-digest \
  --skill web-search --channel latest

Flag	Default	Purpose
<output>	—	Output tarball path (e.g. bundle.tar.gz).
<entity_id>...	—	One or more chain entity_ids to export.
--skill SKILL_ID	—	AgentSkill entity_id to bundle in (repeatable).
--channel NAME	latest	Memory channel to read from.

On success it prints the chain count, skill count, output path, and byte size:

exported 2 chain(s) + 1 skill(s) → bundle.tar.gz (40817 bytes)

If an entity_id can’t be fetched it is skipped and listed (the rest still bundle):

skipped (fetch failed): typo-chain-id

No --json

care export writes a human-readable summary only — there is no --json output. It exits 1 on a bundling error.

care import <PATTERN>...

care import is the inverse of care export: it validates chain JSON files (glob patterns are accepted) and, with --apply, saves them to Memory. It runs as a dry-run preview by default — nothing is written until you pass --apply. See Discovery & validation for the full flag reference.

Flag	Default	Purpose
<PATTERN>...	—	File paths or globs (e.g. **/*.json).
--apply	off	Actually save to Memory (otherwise dry-run).
--channel NAME	latest	Default channel for entries that omit one.

It exits 0 when every chain validates, 1 otherwise.

---

Round-trip: move a library between machines

Export two chains plus the skill they share on the source machine:

care export bundle.tar.gz weather-report risk-digest --skill web-search
# exported 2 chain(s) + 1 skill(s) → bundle.tar.gz (40817 bytes)

Copy bundle.tar.gz to the target machine, unpack it, and preview before committing — the dry-run shows exactly what would be saved:

tar xzf bundle.tar.gz
care import "*.json"            # dry-run preview, writes nothing
care import "*.json" --apply    # commit the chains to local Memory

See also

• Discovery & validation — care import flags and chain validation.
• Memory & Library — browse and curate what you imported.